In March 2008 a founder recorded his screen for about four minutes, narrated it himself, and hid a dozen in-jokes in it. Chocolate Rain. Office Space. XKCD. Then he posted it to Digg and Hacker News. The video looks like a support ticket. Dropbox’s beta waitlist went from 5,000 people to 75,000 in a day.
In November 2023 Humane put out a ten minute film for the AI Pin that was, by every craft measure I was trained to care about, beautiful. In it, the Pin tells someone the April 2024 eclipse will be best seen from Australia. It crossed North America. It says a handful of almonds has 15 grams of protein, which is closer to half a cup. Humane blamed pre-release software and recut the video. By August 2024 The Verge was reporting that the company’s daily returns had overtaken its sales.
Those two videos sit at opposite ends of every quality scale I own, and they produced opposite outcomes in the wrong direction. I make launch films for AI and software companies, so that bothered me enough to go looking properly.
Attention and launch are two different jobs
Here is the thing I think goes wrong before a single frame gets shot. Getting attention and helping a product launch are separate jobs that happen to use the same file format.
Attention is mostly a function of the thumbnail, the first line of the post, the account that published it, the day of the week, and whether somebody with a big following decided to quote it. The film contributes to that. It does not control it.
A launch is a function of whether a stranger finished the video believing something they didn’t believe when they started, and whether the thing they now want actually exists, works, and is available.
Both jobs get scored with the same number. Views. I studied computer engineering before I did any of this, and the habit that stuck hardest is a boring one: when a system misbehaves, check whether you’re reading the right instrument. Views are the instrument everyone reads because it’s printed largest, not because it measures the thing anybody actually needs.
A better film can make the launch worse
This next part took me a long time to accept, because it argues against my own invoice.
A launch film is a promise, and promises are liabilities. The better the film, the higher the bar the product has to clear when somebody finally holds it.
Magic Leap put a whale through a school gymnasium floor in October 2015 and captioned it as shot directly through their technology, no special effects, no compositing. Weta Workshop, the people behind the Lord of the Rings effects, had worked on the company’s concept films. What eventually shipped could not do that. The video didn’t just fail to help. It became the thing people held up as evidence against the company for years afterwards.
Google Glass got better theatre than most products ever get. At I/O in June 2012, skydivers jumped out of a blimp and landed on the roof of the Moscone Center, livestreaming the whole descent through the glasses. The Explorer Edition was $1,500. Google stopped making it in January 2015.
Humane is the cleanest version of the pattern, because the craft was genuinely there. The gap between the film and the device is what everybody ended up writing about.
Nobody prices this risk in. A launch video gets treated as pure upside, a thing that can only help. It can hurt, and the way it hurts is by being persuasive about something that isn’t true yet.
What was actually doing the work
Go back to the Dropbox screencast, because it gets told wrong. The story is usually “a video got 70,000 signups,” which makes it sound like the video did it.
Watch what actually happened. He put it in rooms where people already had that exact problem, engineers who kept a folder on a USB stick and emailed themselves zip files. Then he proved he was one of them, in the only currency those rooms accept, by burying references only they would catch.
The in-jokes weren’t decoration. They were the payload. That video is a positioning move wearing a screencast costume. Same file posted cold to YouTube with a nice grade on it, and nobody tells this story.
Dollar Shave Club is the other one everyone quotes. Around $4,500, shot in a day, launched on 6 March 2012, more than 12,000 orders in the first 48 hours. The servers fell over.
That last detail is the useful one. Servers falling over is the sound of a company that was ready to be paid. There was a subscription, a warehouse and a fulfilment process sitting behind the joke. The film opened a door into a building that existed. Most launch films I get asked about open a door into a construction site.
The statistic everyone quotes backwards
Kickstarter’s own guidance says projects with a video succeed around 50 percent of the time and projects without one around 30. It gets quoted in every deck selling video services, probably including some of mine, as proof that a video helps.
Read it again. It says projects that have videos succeed more often. It doesn’t say the video is why.
The kind of team that makes a video is also the kind of team that wrote a real description, took decent photos, worked out shipping, lined up their first hundred backers before the page went live and knew exactly who they were talking to. The video is a symptom of preparation. Preparation is the cause. Buy the symptom by itself and you get a better looking version of the same failure.
Timing is part of the product
Quibi launched on 6 April 2020, at the precise moment its entire target audience was locked in their homes with a television. The pitch was premium short video for people commuting and queuing. There was no commuting. It had raised about $1.75 billion. It peaked somewhere around 1.1 million subscriber households, announced its shutdown that October, and was gone by December. Roku bought the library for under $100 million.
No amount of film craft fixes that. But I’ve watched teams treat a launch date like gravity when it’s usually a calendar invite somebody made in a planning meeting three months earlier.
The question I ask before I write anything
What does one specific person have to believe at the end of this that they didn’t believe at the start?
Not what do we want to say. Not which features shipped. One person, one belief, one change.
Most briefs can’t answer it, because most briefs arrive as a date. “We launch on the 14th, we need a video.” That’s a deadline, not a communication problem, and when nobody names the communication problem the film defaults to the only thing everyone in the room agrees on, which is the feature list.
Having sat on a few sides of this helps more than it should. If video is all you do, you look at a bad launch and you see a bad video. Once you’ve also designed the site and used the product for a week, you notice how often the video is being asked to do the website’s job, or the positioning’s job, or onboarding’s job. That’s why I use the product before I write a word. Not out of thoroughness. Because the angle is almost always sitting inside the product, in the two minutes where you go, oh, that’s the bit.
Build the film that outlives launch day
Launch attention is rented. Whatever spike arrives comes back down, and faster than anyone plans for.
The film that keeps earning is the one still sitting on the homepage in March, still getting pasted into sales threads, still answering “so what do you guys actually do?” Nobody screenshots that one into the group chat. It just quietly does the job for a year.
So when I’m choosing between the cut that will travel and the cut that will explain, I take explain. Not because travelling is bad. Because a film that explains can also travel, and a film built only to travel has nothing left to do in March.
The uncomfortable version of all this: if the product is genuinely ready and genuinely wanted, a rough four minute screencast can beat a beautiful ten minute film. And if it isn’t, the beautiful film raises a bar you then have to clear in public.
The video is never the launch. It’s just the part of the launch you can see.